Intermetallic Compound Growth in Tin and Tin-Lead Platings over Nickel and Its Effects on Solderability

Solder movement combined with the right time and temperature to decompose NiSnj significantly improves solderability

BY J. HAIMOVICH

ABSTRACT. The growth rates for stable tion. Differential scanning calorimetry control the undesirable effects. Ni-Sn intermetallic compounds (IMC) are measurements determined that NiSn3 Nickel has been successfully used as a much lower than those for Cu-Sn IMC. indeed is a metastable phase that rapidly diffusion barrier to prevent interdiffusion Therefore, Ni appears to be a good transforms into stable IMC's and free tin of Cu and Au. This success prompted choice for a diffusion barrier between Cu at temperatures above the tin melting considerations of using Ni as a barrier and Sn. However, growth of a metasta­ point. The kinetic parameters of NiSn3 between Cu and Sn to prevent growth of ble plate-like IMC is a potential cause for transformation were calculated using Cu-Sn intermetallics and to prolong the long-term solderability degradation. This data from isothermal DSC measure­ shelf life of plated parts. Initially, there IMC has an approximate composition ments, and a time-temperature-transfor­ were some indications that growth rates NiSn3, which does not correspond to any mation (T-T-T) diagram was constructed of Ni-Sn intermetallics are very low. Since of the stable Ni-Sn IMC's on the equilibri­ using these kinetics parameters. The the early seventies, there were several um phase diagram. A long-term, low- implications of the findings on solderabili­ investigations of Ni diffusion barrier per­ temperature aging study confirmed the ty, soldering techniques and accelerated formance for Sn-based platings (Refs. 1- undesirable effects of NiSn3 growth upon aging testing for tin-based platings over 4). The results obtained by these authors solderability. Consequently, the growth nickel underplating are discussed. Also indicated a rather complex behavior. rates for NiSn3 were studied as a function discussed is the work to determine the There are three IMC's in the Ni-Sn of aging temperature, lead content, and actual mechanisms of solderability deteri­ binary system, all stable at room temper­ plating type, and were found to be oration. ature: Ni3Sn4, Ni3Sn2 and Ni3Sn (Ref. 5). affected by all of these variables. Lead Early work indicated that out of these was determined to reduce the NiSn3 Introduction three intermetallics, only Ni3Sn4 is growth in matte Sn-Pb. The growth rate present. The growth rates of Ni3Sn4 at reaches a maximum between 100° and The growth of intermetallic com­ 70°C (158°F) were found to be much 140°C (212° and 284°F) and then pounds in various platings affects solder­ lower than those of Cu-Sn intermetallics decreases. This behavior is indicative of a ability, and is one of the main factors in (Refs. 2, 3), and our results confirm this — metastable phase, and so is the composi- determining the shelf life of plated com­ Fig. 1. However, together with the slow ponents. It has been known for several growth of Ni3Sn4, an extremely fast decades that the formation of certain tEHPERATURE, *C growth of an intermetallic of unknown 250 150 100 50 20 intermetallic compounds (IMC's) in tin nature was detected (Refs. 1, 3). The W3 F and tin-based platings can have an unde­ composition of this intermetallic roughly sirable effect resulting in a serious corresponds to NiSn3, which cannot be degrading of component solderability. identified with any of the intermetallics Consequently, a considerable research on the phase diagram. effort was undertaken by the plating The characteristic platelet morphology industry to understand the formation and of this compound is shown in Fig. 2. It has growth of IMC's, and, if possible, to been suggested that the extremely fast growth of this intermetallic can cause deterioration of solderability (Ref. 3). This happens when the platelets penetrate all KEY WORDS the way through the tin layer to the surface and then oxidize. Therefore, Intermetallic Growth from the practical point of view, it is Tin/Tin Lead Plating important to be able to control this unde- Platings on Nickel Solderability Testing Low-Temperature Aging Long-Term Aging /. HAIMOVICH is a Development Engineer in NiSn3 Decomposition the Materials Engineering and Research Divi­ NiSn3 Intermetallic sion, AMP, Inc., Harrisburg, Pa. NiSn3 Metastability Fig. 1 — Intermetallic growth in matte Sn on Cu Thermal Stability Paper presented at the 12th Annual Electronics and Ni substrates. Solid line — Ni; dotted line — Manufacturing Seminar, held February 18-20, Cu 1988, at China Lake, Calif.

102-sl MARCH 1989 sirable growth.

At the start of our research on NiSn3, the knowledge about the compound was rather limited. We knew that it grew in significant amounts only at lower temper­ atures, below approximately 160°C (320°F). Above this temperature, only a continuous layer of Ni3Sn4 had been observed. NiSn3 grew in a large variety of platings. The growth of the compound was the largest in bright tin over bright nickel platings, and the lowest in matte Sn over sulfamate Ni (Ref. 3). Another study found that it was not possible to control the occurrence or growth rate of the compound by variation in plating param­ eters, both for tin and nickel. The occur­ rence of the compound was either elimi­ nated or greatly reduced in nonplated layers (Ref. 6). Fig. 2 — Intermetallic growth in matte Sn, Ni underplate. Aged 1.5 years at 50°C. A —X1000; This paper can be divided into four B-X5000 parts. The first part is a brief description of a long-term, low-temperature solder­ These results contradict to some extent layer, Ni3Sn4, and discontinuous "nee­ ability study for matte tin over Cu and the results of the first test, since platings dles" (actually, platelets), NiSn3. The addi­ electroplated Ni. The second part with and without underplating display tion of 10% Pb greatly reduces the presents results of our study of short- solderability of the same order. amount of NiSn3 and the addition of 40% term growth rates of NiSn3. The third part To investigate the cause of the solder­ Pb practically eliminates it. This correlates describes the work on the thermal stabil­ ability failure, all samples were cross- well with dip test results. Most likely, the ity of NiSn3. And finally, we will discuss sectioned to reveal IMC growth. Such thin layer of Ni3Sn4 does not affect the how what we learned about this interme­ cross-sections for matte Sn, 90% Sn-10% solderability at all, and all the detrimental tallic was applied to the determination of Pb and 60% Sn-40% Pb (ail over Ni effects come from NiSn3, particularly if it actual mechanism of solderability deterio­ underplating) are shown in Figs. 3A, B and grows all the way to the surface or close ration and ways to improve solderabili­ C. For matte Sn and 90% Sn-10% Pb, the to the surface. Another interesting phe­ ty. intermetallic consists of a thin continuous nomenon that was found in this study

Long-Term, Low-Temperature Aging Solderability Study Table 1—Solderability Test Results (% Passed)

The purpose of the study was to assess Aging Time the effects of low-temperature aging on Sample 6 Months 12 Months 18 Months 24 Months solderability of matte tin and tin-lead platings over copper and electroplated A-Aged at 50°C nickel. The variables were: 1) composi­ 100 Sn/Ni 100 57 43 0 tion-100% Sn, 90% Sn-10% Pb, 60% 90 Sn-IOPb/Ni 100 86 86 86 Sn-40% Pb (all wt-%); 2) temperature- 60 Sn-40Pb/Ni 100 100 100 100 24°C (75°F) (room temperature) and 100 Sn/Cu 86 86 29 0 90 Sn-19Pb/Cu 38 38 0 0 50°C (122°F) (maximum warehouse tem­ 60 Sn-40Pb/Cu 29 29 14 14 perature); 3) time — samples taken out for testing at 6 month intervals, up to 24 B-Aged at 24'C 100 Sn/Ni 100 months. Plating thickness was 200/x in. ± 100 75 0 90 Sn-IOPb/Ni 100 100 86 86 50, underplating (Ni) thickness was about 60 Sn-40Pb/Ni 100 100 100 100 100M in.; substrates were pure copper. 100 Sn/Cu 100 86 43 14 Solderability testing was done by two 90 Sn-IOPb/Cu 14 14 14 0 methods. First is the test procedure for 60 Sn-40Pb/Cu 75 50 25 25 estimating solderability of metallic sur­ faces, a dip test, using nonactivated rosin flux. The required coverage to pass the test is 95%. Table 1 represents results of Table 2—Meniscograph Test Results (from Ref. 7) the test. We can see that tin-lead over nickel has the best solderability. These Aging Time 24 Months results show a trend, but are somewhat Aged at 24 °C Aged at 5C °C subjective, since they depend on the Final Value Time to Final Value Time to operator's judgment. Therefore, a sec­ of Wetting Cross Zero of Wetting Cross Zero Sample Force (g) Force (g) ond method was used, the menisco- (s) (s) graph, or wetting balance (Ref. 7). Two 100 Sn/Cu -0.23 ± 0.02 — -0.29 ± 0.03 — characteristics were employed to evalu­ 100 Sn/Ni -0.14 ± 0.03 - -0.20 ± 0.03 - ate solderability: the final value of wetting 90 Sn-10 Pb/Cu 0.05 ± 0.01 2.9 ± 0.4 0.05 ± 0.01 3.1 ± 0.3 force and the time to cross zero force 90 Sn-10 Pb/Ni 0.05 ± 0.01 2.7 ± 0.2 0.02 ± 0.01 3.9 ± 0.6 line. Results for samples aged for two 60 Sn-40 Pb/Cu 0.18 ± 0.05 2.1 ± 0.2 0.16 ± 0.06 2.2 ± 0.3 years are presented in Table 2 (Ref. 7). 60 Sn-40 Pb/Ni 0.14 ± 0.03 2.2 ± 0.2 0.13 ± 0.03 2.3 ± 0.1

WELDING RESEARCH SUPPLEMENT 1103-s Fig. 3 —Decrease of Ni-Sn IMC growth with increasing lead content. Fig. 4 —Decrease of Ni-Sn IMC growth with high-temperature prean- A - 100% Sn;B-90% Sn-10% Pb; C-60% Sn-40% Pb. Aged 1.5 years nealing. A — No preanneal; B — was preannealed at 155 "C for 8 hours. at 50 C (1000 X) All aged at room temperature for 1.5 years (1000 X)

was the effect of preaging at a higher ing the time frame for the aging tests. The and 40 ASF. Then half of the substrates temperature. Figures 4A and B show maximum growth rates can be used for were plated with matte Sn, and the other decrease in the NiSn3 IMC growth for accelerated aging tests. half with 75% Sn-25% Pb. Matte Sn was 100% Sn samples. This decrease is the The variables in our test were aging plated using a stannous sulfate bath at result of preaging at 155°C (311°F) for temperature and composition of the tin room temperature and 40 ASF, and matte eight hours. Samples were consequently alloy plating. The temperatures varied Sn-Pb was plated using a fluoborate sys­ aged at room temperature for 1.5 years. from 75° to 225°C (167° to 437°F). tem at room temperature and 40 ASF. Figures 5A and B show the same effect Isochronal anneals at three temperatures The baths used were standard baths for the 90% Sn-10% Pb sample; preaging were performed at a single time, with from the manual plating line. nearly eliminated subsequent NiSn3 annealing times starting from 16 days for All aging treatments were performed growth. 75°, 100° and 125°C anneals, and down in Blue M air circulating ovens. The aged to 2 days for 210° and 225°C anneals. samples were analyzed in two ways. These times were chosen to allow suffi­ Study of NiSn Growth Rates Some of them were mounted, cross- 3 cient growth of intermetallic at appropri­ sectioned and etched to reveal the inter­ The purpose of this study was the ate temperatures, so that the growth metallic. The others were stripped of determination of the growth rates of the rates at various temperatures can be remaining free tin in the International Tin NiSn3 compound from 75°C (167°F) up easily compared. Research Institute (ITRI) stripping solution to the melting point in matte Sn and 75% Substrates suitable for metallographic for consequent examination by optical Sn-25% Pb (wt-%) platings, and in particu­ examination and scanning electron and scanning electron microscopy. Tin lar, the temperature region in which the microscopy were prepared from copper stripping is much less time consuming growth rates are at their maximum. strip. They were plated with Ni using a than cross-sectioning and, therefore, can This objective is important in establish­ sulfamate bath operated at 60°C (140°F) be used for fast analysis of a large num-

C

g. 5 — Combined effect of preannealing and lead content; 90% Sn-10% Fig. 6—Matte Sn on Ni; tin etched. Aged 76 days at: A —75°C;B— 100° Pb. A—No preanneal; B~was preannealed at 155"C for 8 hours. All C; C- 125 "C (1000 X) aged at room temperature for 1.5 years (1000 X)

104-s| MARCH 1989 Fig. 7-Matte Sn on Ni; tin removed. Aged 8 days at: A- 135°C; Fig. 8-Matte Sn on Ni; tin removed. Aged 2 days at: A-165"C; B- 145°C; C- 155°C (1000 X) B- 175°C; C- 185"C (1000 X)

ber of samples. disappearance of NiSn3 after aging above Both cross-sectioning and tin-stripping 200°C. This disappearance can be produced similar results for matte Sn explained by the second cause, the insta­ samples. The size of the platelets bility of NiSn3 at these temperatures. As C increases up to 125 C (257°F) (Fig. 6) will be described below, NiSn3 is not a then stays relatively constant up to thermodynamically stable phase. At ele­ V5»i*' ^MMWMJMW 155°C (311°F) (Fig. 7) and then vated temperatures it transforms by decreases —Fig. 8. The number of plate­ decomposing into free tin and stable lets continually declines with rising tem­ tin-nickel IMC's. perature. No platelets were observed in cross-sections above 175°C (347°F). The tin stripping method is more sensitive: Study of Decomposition single platelets are still observed at tem­ Parameters of NiSn3 peratures as high as 195°C (383°F)-Figs. 9 and 10. It is likely that crystals observed Thermal Stability of NiSn3 at high temperatures nucleated and par­ The transformation of NiSn3 at higher tially grew at room temperature during temperatures was studied by differential the time that passed between plating and scanning calorimetry (DSC). The intent annealing (several months), and then has been to characterize the thermal went through fast growth during heating stability of the NiSn3 IMC. up to the aging temperature (several The weight of a DSC sample has to be minutes). The orientation of platelets anywhere from 1 to 10 mg. Under nor­ seems to be random. mal aging conditions (i.e., single-tempera­ In the case of matte 75% Sn-25% Pb ture aging, reasonable aging time), the plating, the number and size of the plate­ amount of intermetallic in aged plating lets were not large enough for the plate­ will not be sufficient for DSC work. lets to be observed in cross-sections by Therefore, a special two-step aging was optical microscopy. First of all, this con­ used based on our previous findings firms our previous finding that lead dras­ about NiSn3 growth rates. First, samples tically decreases both the density of the were aged at 75°C for 2 months to platelets and the rate of their growth. nucleate and start growing NiSn3 crystals. Otherwise, the growth is very similar to Second, they were aged for 1 to 2'/2 that in matte Sn. The size of crystals months at 125°C to increase the crystals' increases and density decreases up to size to maximum. Two kinds of samples 125°C-Fig. 11. Then the amount of were aged in such manner: matte Sn over intermetallic stays relatively constant up sulfamate Ni underplating and bright Sn to 145°C, and drops at 155°C-Fig. 12. over bright Ni underplating. The second sample's choice was based on work that The total mass of NiSn3 IMC, which is a reported the highest NiSn growth in product of the density of platelets and 3 such samples (Ref. 6). The first sample their average size, will peak at some produced a large quantity of IMC; the intermediate temperature, most likely concurrent growth of stable IMC was between 100° and 125°C. This behav­ minimal —Fig. 13. The second sample ior has two probable causes. First, if NiSn 3 (bright Sn) produced amounts sufficient nucleates at the Ni-Sn interface at a for DSC measurement, but not as large, higher temperature, it has to compete and concurrent growth of stable IMC Fig. 9—Matte Sn on Ni; tin removed. Aged 76 with nucleation and growth of Ni Sn . But 3 4 was substantial. It was found afterward days at: A-75°C; B-100°C; C-125°C this does not account for the complete (1000 X)

WELDING RESEARCH SUPPLEMENT 1105-s that the second step temperature (125°C) was too high for that sample — aging below 100°C produces much bet­ ter results. The aged samples (coupons 1X1 in.) were stripped of free tin using the ITRI stripping solution. The DSC samples were prepared by carefully scraping the IMC from the substrates into the DSC's alumi­ num sample pans.

A typical DSC scan for the NiSn3 inter­ metallic is shown in Fig. 14. The sample, in this case, was grown in matte Sn over sulfamate Ni. Figure 15 is a typical scan of decomposed IMC sample (i.e., second scan). A pure tin sample scan is used as a standard to calibrate temperature. The onset of the peak (i.e., melting point) is about 503 K. The actual melting point of tin is 505 K, so the temperature scale must be corrected by 2 deg for this particular scan rate, 40 K/min. (In the following discussion, all temperatures are corrected.) A DSC scan for "matte Sn" IMC sample (Fig. 14) has a peak with onset at about 513 K, 8 deg higher than melting point of tin (in Figs. 14 through 18, the corrected temperature is in parentheses). The difference between the tin melting point (standard) and the IMC peak onset decreases with decreas­ ing scanning rate. Figure 16 shows a scan for NiSn3 at a lower scanning rate. The difference between tin melting point and onset is about 3 deg. The shift of the peak to lower temperatures with Fig. 10-Matte Sn on Ni; tin removed. Aged 2 days at: A- 125°C; B- 145"0 C- 165°C- D-185°C(1000 X) decreasing scanning rate is characteristic of a metastable phase that transforms by a thermally activated process. ",> The IMC NiSn3 decomposes into tin x m and stable intermetallic compounds. This |^ v \V.'"< was confirmed by x-ray diffraction of the decomposed sample, which detected mostly tin and also Ni3Sn2 and Ni3Sn4 (Ref. 8). It was also confirmed by SEM micro­ . 5 H probe that detected Sn and Ni3Sn4 only. Therefore, the peak observed upon sec­ ' *^^%*^ ond and following scans is produced by tin that resulted from NiSn3 decomposi­ tion—Fig. 15. However, the onset of the peak in Fig. 15 (at 503.2 K) is 1.9 deg lower than for pure tin at the same Fig. 11-Matte 75% Sn-25% Pb on Ni-solder removed. Aged 16 days at: A-75°C; B- 100°C scanning rate. There are two likely rea­ (1000 X) sons for this shift. First, the product of decomposition is not pure tin but tin- nickel eutectic. The melting point of this eutectic is 504 K, 1 deg below that of tin. The observed melting point is shifted down by about 1 K more because the sample is in the form of fine powder, so the surface energy component of the free energy becomes important. Figure 16 also illustrates another fea­ ture of NiSn3, a small peak before the IMC peak. The onset for this peak is at 503.2 K, similar to that of the fully decom­ posed sample in Fig. 15. Hence, that peak was attributed to tin that evolved as a Fig. 12-Matte 75% Sn-25% Pb on Ni, solder removed. Aged 8 days at: A - 125°C; B- 155°C result of partial decomposition of NiSn3 (1000 X) below 503 K. This assumption was

106-s I MARCH 1989 2 WT. 1. 10 ng a. SCAN RATE. 40. 00 dag/m O _i UJ PEAK FROM, 50B. 48 IAX, 518.21 > TO. 529. 26 ONSET. 511.23 CAL/GRAM, 8. 27 X o oc < ui Ui

z 2UJ a. 510.M 520.00 530.00 O Fig. 13 —Intermetallic growth in matte Sn over sulfamate Ni, aged 2 months at 75°C, and consequently 1 month at 125°C. Free tin removed TEMPERATURE (K) (1000 X) Fig. 14-A typical DSC scan for NiSn3; matte Sn sample. Scan rate 40 deg/min. Onset at 511.3 (513.3)K. The corrected temperature for this and the following figures through Fig. 18 is given in parentheses O tr < Ul Ui WT. 6. 10 mg MAX! 503. 68 SCAN RATEi 40.00 dog/min 1 WTi 2.40 mg I- SCAN RATE. 10.00 dag/mln PEAK FRO* 487 Z TOi 511.33 ui ONSET, 501. 17 CAL/GRAH. 8. 12 2 ) PEAK FROM, 4SS. 04 TO. 515.96 0. ONSET. 503. IB CAL/GRAM. 7.96 7.50 o

X o oc < UJ / ' 1' . Ui 0.00 '.'•a. co ita. oc .do. oo 500.00 510.00 520.00 530.00 5*0.00 TEMPERATURE (K) z Fig. 15 —A typical DSC scan for decomposed NiSn; matte Sn sample. UJ TEMPFRAT'JRE CK) 2 Scan rate 40 deg/min. Onset at 501.2 (503.2)K3 a. O Fig. 16 — DSC scan for bright Sn sample; lower scan rate, 10 deg/min. -i Onset at 503.2 (508.0)K >ui

O WT. 2.30 mg CC SCAN RATE. 10. 00 d WT, 2.30 mg < SCAN RATE. 10.00 dog/ UJ PEAK FROM. 497. 57 Ui TO. 499. 82 fJNSET, 4EK 24 PEAK FROM. 496. 17 CAL/GRAM. 02 TO. 500. 96 ONSET. 498.28 CAL/GRAM. 1.S6 z UJ s 0- O

X u cc < Ul Ui TEMPERATURE (K) UJ

Fig. 17 - First scan for NiSn3 sample. Scan was stopped at 506 (510.8)K, oc and sample cooled. Tin peak at 499 (503.8JK is very small, since only a Fig. 18 — Second scan for the sample in Fig. 17. Tin peak is much larger small fraction of NiSn3 decomposed below 498 (502.8)K because NiSnj partially decomposed when heated to 506 (510.8)K

WELDING RESEARCH SUPPLEMENT 1107-s 1 2 LOG(TIME),sec

LOGCTIME),sec

Fig. 19 - Isothermal DSC curves. The sample is NiSn3 that grew in matte e I Sn over sulfamate Ni during 75°C/2 months and 125°C/1 month, LOGCTIME),sec two-step aging Fig. 20—Avrami plots obtained using data from Fig. 19 checked by scanning the temperature where x is the fraction of material trans­ n and k can be measured directly from part way into the IMC peak and thus formed at the time t, n is a dimensionless the plot: n is the slope of the plot, and partially decomposing the intermetallic exponent, and k is the reaction rate; k nlnk is the intercept. The DSC technique (Fig. 17), then cooling the sample and changes with temperature exponentially does not measure fraction x directly; performing the second scan all the way (Arrhenius type of relation): instead, dx/dt (the x derivative with time) through the IMC peak —Fig. 18. It can be is measured. The measurement produces a curve as those in Fig. 19. The trans­ seen that the first partial scan resulted in a k = k0exp i (2) large increase of the first peak, thus RT formed fraction x at time t can be deter­ mined as the ratio of area under the confirming that the first peak is the tin where E is the activation energy, k the 0 curve up to time t (hatched) to the total peak. frequency factor, R the gas constant and area under the curve. If this fraction x is T the temperature (Kelvin). The Equation 1 can be rewritten as: 1 Kinetic Parameters plotted as Inln (• vs. Int, we obtain of NiSn3 Decomposition 1 1 -x Inln ( ) = nlnt + nlnk (3) plots as in Fig. 20. It is linear at lower 1 -x Most isothermal reactions in solid values of t, but deviates from linearity at metallic phases are described by the Therefore, if x vs. t is plotted in coordi­ higher times indicating a lower transfor­ Mehl-)ohnson-Avrami (M-J-A) equation mation rate than that expected according (Ref. 9): i to M-J-A equation. This nonlinearity at n nates Inln (" vs. Int, the parameters x = 1-exp [-(kt) ] (1) i high values of t is most likely due to the

108-s | MARCH 1989 TEMPERATURE, K 520 510 500 490

OX) 1 - r

-2 -

^\ -* r

-" r LOG(TIME),sec

"5

Fig. 21-Arrhenius plot for rate constant k. Circles are data points; line is fit through data points restriction of crystal growth by the small Fig. 22- size of the analyzed NiSn3 particles (Ref. A — Sigmoidal curves 10). It must be noticed that as a rule, the for decomposition nonlinearity becomes significant at of NiSnj. The circles around x = 0.8, i.e., when the 80% of are the actual data, NiSn3 is decomposed. Therefore, the use 0 1 and the curves are of M-J-A equation is still justified. After LOGCTIME),sec A vrami fits using the kinetic parameters are determined, n = 1.5 and k we calculate the fraction of transformed obtained from Fig. material as a function of time at isother­ 21. The mal conditions —Fig. 23. From that, the temperatures indicated are in time-temperature-transformation (T-T-T) degrees C. Sample diagram can be calculated — Fig. 24. aged 75°C/2 Samples. The sample used in this study months and was NiSn3 grown in matte Sn plated over 125°C/2 months; sulfamate Ni. A substantial amount of the B — Same as Fig. 22 sample was produced by two-stage A, but for sample aged 75°C/2 aging. Unfortunately, there was not months and enough of the original sample aged at 125°C/2.5 months 75°C for 2 months and, consequently, at C — Same figure as 125°C for 2 months. Therefore, another 22 A, but for sample sample aged at 75°C for 2 months and at aged 75 "C/2 125°C for 2.5 months was used to months and extend the temperature range of mea­ LOGCTIME),sec 125°C/1 month surements from 517 to 521 K. Another similar sample, aged at 75°C for 2 months thermal DSC is the initial instability are shown in Fig. 20. The Avrami plots and then at 125°C for 1 month, was used because of high heating rate prior to the were used to calculate parameters n and to confirm the results of the first series of start. This instability was partly overcome k. The average value of the parameter n measurements. The weights of samples by using larger samples. The very initial is 1.5. The reaction rate k is plotted in Fig. used in this investigation were between 2 part of the DSC runs was determined by 21. It has an Arrhenian dependence, as and 5 mg. extrapolation to the baseline at time expected. Using the determined n and k, Isothermal DSC. The differential scan­ t = 0. The data were stored and then the transformed fraction was calculated. ning calorimetry (DSC) is not normally plotted out in an appropriate scale. The Figures 22 A, B and C show the calculated used for isothermal measurement. How­ plots were digitized using an HP digitizer, decomposition curves together with ever, our DSC apparatus can be set up and then stored for further analysis. data; the fits are very reasonable. Figure for isothermal runs. Usually, the specimen The typical isothermal DSC curves are 23 is the same plot, but for a wider range was heated up to about 40 K below the shown in Fig. 19, which covers the whole of temperatures, using extrapolated val­ temperature of isothermal run and equili­ range of temperatures used. Using these ues of k from Fig. 21. And finally, Fig. 24 brated at this temperature. Then it was data, the fraction x was calculated and presents a T-T-T diagram calculated using heated as fast as possible to the temper­ then plotted in coordinates log- Fig. 23. The left boundary of the T-T-T ature of the run (anywhere from 505 to diagram represents the beginning of 521 K). Data acquisition was started log ( ) vs. log t, where the loga­ transformation (5% decomposed), and immediately after reaching the working 1 -x the right boundary, the completion of temperature. The main problem in iso- rithm with base 10 is used. These plots transformation (95% decomposed).

WELDING RESEARCH SUPPLEMENT 1109-s 0 2 3 LOGCTIME),soc LOGCTIME),sec

Fig. 23 — Sigmoidal curves for decomposition of NiSnj. Obtained in Fig. 24 - T-T-T diagram for NiSn3 IMC. The line on the left represents the same manner as Fig. 22 A beginning of transformation (5% transformed); the line on the right, the completion of transformation (95% transformed)

Metastable IMC: actual mechanism of solderability deterio­ decomposition cause a lot of surface Engineering Implications ration, and ways to overcome the deteri­ irregularities. oration. Based on work to date, some Indeed, raising the temperature to Effects on Solderability trends are already evident. 275°C did not substantially improve the As we can see from the T-T-T diagram surface appearance. The increase of dip Since the stable IMC Ni3Sn4 grows very (Fig. 24), NiSn3 will fully transform at time from 5 to 30 s (to allow the decom­ slowly at room temperature, the deterio­ 260°C (470°F) in less than one second, position products to float off) had only a ration of solderability in tin platings with while at 215°C (419°F) it will take about marginal effect. The best results were nickel underplating is caused by fast 10s seconds (i.e., about 28 h). Samples achieved by the combination of time and growth of metastable NiSn3. The exact were aged (using two-step aging) to temperature adequate for nearly full mechanism of deterioration is not clear at produce large intermetallic growth and decomposition of NiSn3 (10 s at 245°C- the moment; the most likely cause is then tested for solderability at these tem­ Fig. 24), combined with sample move­ oxidation of NiSn3 near or at the surface peratures (5 s dip). The 215°C samples ment in the solder, to facilitate the of the plating (Ref. 3). The metastability of have large nonwetted areas, and the removal of decomposition products. Fig­ NiSn3 might offer an opportunity to offset wetted areas display a lot of surface ure 27 shows these test samples; move­ the undesirable effects by adjusting the irregularities —Fig. 25. Wetting is much ment of the samples markedly improved process variables. The amount of stable better for 260°C samples, but again there the regularity of the surface. The repre­ intermetallic after NiSn3 decomposition is are a lot of surface irregularities (bumps). sentative cross-sections are shown in Fig. much less than the amount of original The cross-sections of the test samples 28. The IMC is nearly absent in the NiSn3. For example, 1 g of NiSn3 will reveal that these bumps are caused by samples that were moved. decompose into 0.2 g of Ni3Sn2, 0.2 g of needle-like IMC in the solder layer —Fig. Based on these tests, it is possible to Ni3Sn4 and 0.6 g of tin. Therefore, the 26. The IMC compositions were analyzed predict how NiSn3 intermetallic will affect soldering process that exposes the inter­ by microprobe. For 260°C, the composi­ solderability in various soldering pro­ metallic to high enough temperature for a tions correspond to Ni3Sn4; NiSn3 is defi­ cesses. In wave soldering, at a solder sufficient time to decompose NiSn3 might nitely absent. In 215°C samples, the IMC wave temperature of 255°C (491 °F), it result at least in a partial restoration of is NiSn3. Thus, just raising the tempera­ will decompose in less than a second, and solderability. Below, we will briefly ture is not enough to insure good solder­ at 265°C (509°F), in less than 0.1 s.; this is describe the work on determining the ability, since the products of NiSn3 combined with the movement of the

Fig. 25-Solderability test samples tested at 215"C (top) and 260"C Fig. 26 — Cross-sections of the solderability test samples tested at 215"C (bottom) (top) and 260°C (bottom)

110-s|MARCH 1989 •Hi

Fig. 27—Solderability test samples tested at 245 "C top, stationary; Fig. 28 — Cross-sections of solderability test samples tested at 245°C, bottom, moving top, stationary; bottom, moving

solder. On the other hand, at the vapor after a relatively short time. This deterio­ editing this paper, and to M. L. Smith for phase reflow temperature, 215°C, it will ration is not caused by the growth of typing it. I want to thank E. Carver, E. take about 28 h to fully decompose the stable Sn-Ni IMC's, which is extremely Walker and C. LaRosa for help with DSC. intermetallic. Therefore, the presence of slow, but by the fast growth of a meta­ I am also thankful to W. Hadley for help NiSn3 might be acceptable in the case of stable NiSn3 IMC. The NiSn3 grows in the with data digitizing and HP program­ wave soldering, but will be unacceptable form of platelets. The number of platelets ming. for vapor phase reflow. per unit area continually decreases with increasing temperature. The average size Accelerated Aging reaches a maximum around 125°C and References decreases above 155°C. The addition of The growth rates of NiSn3 IMC go 1. Kay, P. )., and Mackay, C. A. 1976. The 10% Pb greatly reduces NiSn3 growth, growth of intermetallic compounds on com­ through a maximum at temperatures and the addition of 40% Pb practically mon basis material coated with tin and tin-lead between 100° and 140°C, depending on eliminates it. alloys. Trans. Inst. Met. Fin. 54: 68. the plating's type. This fact, coupled with It was determined by differential scan­ 2. Kay, P. J. and Mackay, C. A. 1979. Barrier the metastability of the NiSn3 IMC, dic­ ning calorimetry (DSC) that NiSn is a layers against diffusion. Trans. Inst. Met. Fin. tates caution in the use of conventional 3 metastable phase, which decomposes 57: 169. accelerated aging. The high-temperature into free tin and stable Sn-Ni IMC's. Iso­ 3. Harman, A. C. 1978. Rapid tin-nickel aging, above 140°C, may not be used at intermetallic growth: some effects on solder­ thermal DSC was used to calculate the all, since it will not detect the potential for ability. Proceedings, InterNepcon. Brighton, kinetic parameters of the NiSn3 transfor­ U.K. p. 42. NiSn3 growth at lower temperatures. mation, and a time-temperature-transfor­ 4. Lindborg, U., Asthner, B., Lind, L, and The morphology of growth (particular­ mation (T-T-T) diagram was constructed Revay, L. 1975. Intermetallic growth and con­ ly density of growth) at temperatures using these parameters. This diagram pre­ tact resistance of tin contacts after aging. Proc. higher than 100°C may differ from that at dicts that while high-temperature solder­ 21st Ann. Holm Seminar on Electrical Contacts. room temperature or slightly above. Due ing processes, such as wave soldering, Chicago, III., p. 25. to complexity of the situation, a new might not be affected by the presence of 5. Hansen, M., and Anderko, K. 1958. approach such as two-step aging may be Constitution of Binary Alloys. McGraw-Hill, NiSn3, the processes that use lower tem­ appropriate. This aging will be similar to peratures, such as vapor phase reflow, N.Y. the aging used to grow a DSC sample. will be affected. The testing of aged 6. Harman, A. C, and Park, C. A. 1980. Rapid intermetallic growth in tin and tin-lead During the first, low-temperature step samples showed that the combination of (anywhere from 50° to 100°C), small coatings on nickel underlayers. Report No. time and temperatures adequate to IR/R446/1986/938, Standard Telecommunica­ crystals of NiSn3 will nucleate and start to decompose NiSn3 and solder movement tions Laboratories, Ltd., Harlow, Essex, grow. During the second, high-tempera­ can drastically improve solderability. England. ture step (from 100° to 140°C), these 7. Davis, T. AMP Incorporated, Unpub­ crystals will undergo fast growth. It was lished Research, 1985. already shown that such aging can 8. Kahn, D. AMP Incorporated, Memoran­ Acknowledgments produce NiSn3 with both large crystal size dum to J. Haimovich, May 28, 1986. and high growth density. This kind of IMC I am grateful to all the people at AMP 9. Augis, J. A., and Bennett, J. E. 1978. growth is similar to that produced during ME & R who made this work possible: to Kinetics of transformation of metastable tin- long-term natural aging. E. Carlevale for preparing numerous nickel deposits. 1— determination of the cross-sections and for performing solder­ Avrami equation parameters by DSC or DTA. /. Electrochem. Soc. 125:330. ability tests; to E. Carver for helping with Conclusions 10. Speyer, R. F., and Risbud, S. H. 1983. tin stripping and for doing powder x-ray Methods of determination of the activation Solderability of matte tin platings over analysis; to M. Smith for plating the sam­ energy of glass crystallization from thermal nickel underplating, stored at room tem­ ples; to R. Geckle for SEM and micro­ analysis. Physics and Chemistry of Glasses 24: perature or slightly above, deteriorates probe work; to J. Hoyt and G. Lurie for 26.

WELDING RESEARCH SUPPLEMENT 1111-s